|
Any progress with GPU version?
|
|
|
|
There are hundreds of SRBMiner users without any problems, where the miner works flawlessly.  Or they just didn't write about them... Many added deleting of .srb files into start.bat... Please, don't get mad on me... I just think that all next AMD drivers will be with the same features as 18.4.1, 18.5.1 and 18.5.2. What about duplicate shares? |
|
|
|
I notice on 1.5.8 often appears duplicate shares on nicehash heavy algo. If it's some way to reduce amount of duplicate shares or fix that?
Are you sure those are duplicate shares? There is a mechanism that does not allow sending of the same result twice. Also for nicehash there is a protection for stale shares, cause nicehash does not like them. Your miner show that it was finded rejected duplicate share. And on statistic report on "s" key it start to show: Errors: Duplicate share.: 1 |
|
|
|
But I think that doctor must update miner to support latest AMD drivers without issues.
Serious miners don't always update their drivers when a new one comes out, they use ones that are proven to be good and stable. A driver that isn't capable of recognising its own GPU is a trash, and shouldn't be used at all. Altough no one else except you reported this issue, so i guess it's something on your side. I really can't check out every new driver and re-work miner every time when a new driver comes out. You should stick to a driver that is working stable and good for you. I know it all. But I was forced to change drivers because of Windows 1803 update. Only on 18.4.1 driver was added WDDM 2.4 support, but it was too buggy. I used sometime 18.3.4 on 1803, but sometimes I encountered bugs on Windows start and work. When 18.5.1 driver was released I installed him and it work in all miners. But on some of them it's wrong GPU detection. All other mining functions work. And you as serious developer of miner with devfee! must update your miner to modern conditions in wich miners are placed. Creating .srb files on each start - it's the problem, isn't it? Unstable speed of mining is second problem as I know. Many of users of your miner finds different methods to force miner to mine with max speed. Anyone used GPU-Z launching during mining, others just relaunch miner many times, untill miner gives max speed. |
|
|
|
|
I notice on 1.5.8 often appears duplicate shares on nicehash heavy algo. If it's some way to reduce amount of duplicate shares or fix that?
|
|
|
|
Everyone with this problem uses 18.5.1 drivers ?
I use dag fix beta drivers, and have same issue... Only with V1.5.8 ? I have this issue with 18.5.1 and 18.5.2 on 1.5.7 and 1.5.8 versions. Windows 10 1803 x64. And wrong names of cards. 1st 2 RX 580 cards detects as R9 200 series. I have 3rd card r9 270X. |
|
|
|
Hi Dok
V1.5.8 on startup, it generates the .srb file again
This is normal?
It should be generated once for every different gpu, after that on miner start that file should be read, not created again. Hi Doktor, I think this could be better communicated by the program. It *always* says "compiling" although it's just reading the cached / pre-compiled one. If the program finds a pre-compiled srb file, it could say "loading pre-compiled library" or something like that instead of compiling. It happens on 18.5.1 and newer drivers. SRB miner wrong reads cards IDs. Because of it each start it have new ID's of cards and creats new kernel bins. Look at names of .srb files. After name of card it place long string of numbers before end. And each time this numbers are differes. It's exists 1 method to prevent this. But it's not a solution. Disable compute mode on Radeon Settings and than enable it again for each card. But I think that doctor must update miner to support latest AMD drivers without issues. |
|
|
|
|
Claymore, please add heavy algo support. You can return the fee and update miner!
|
|
|
|
|
Zawawa, please, return button "Release memory". I really need them! You may make them not on main page...
|
|
|
|
|
Version 1.5.6. Windows 1803 with 18.5.2 (18.5.1 same thing). Miner didn't release all video memory on closing miner. I use GPU-Z to see vmemory usage. On each RX 580 card after closing miner still exists used vmemory. And with every launch free amount of vmemory decreases. I don't know how, but after many-many launches, during mining GPU-Z shows 15-16Gb of used video memory on each card. But they have only 8Gb of it...
Pagefile usage grows with that too...
Thats definitely something wrong with SRB miner on last Windows version with last AMD drivers...
And still miner detects RX 580 cards as R9 200 series. I have 3rd connected card - R9 270X. Maybe thats the cause of wrong detection of 580?
|
|
|
|
|
Please, make on GPU version video memory usage indicator to user know how much vmemory he allready used for mining.
|
|
|
|
|
And check "Relaunch" option. For me it just closes GGS. After running manually again GGS start mining without questions, so as it continues relaunching...
|
|
|
|
No, only 1.5.7 version... Tested...
yes i decided finally to put a virus into it, now is time to steal all your pictures from my documents  I don't think you will be glad to see them ))))))) |
|
|
|
how should I enter the command that adl is enabled?
If you don't point option to disable adl it enabled by default. |
|
|
|
On AMD 18.5.1 driver on 1803 Windows 10 x64 my RX 580 recognized by SRB miner as R9 200 series [Ellesmere] and every launch build new kernels... It's abnormal. I think, you must update miner to support new drivers, doctor...
By the word, all Claymore's miners detects my cards right on 18.5.1...
And I still can't understand, why every lounch of SRB miner speed is different from each other... To reach max speed it's need to launch SRB several times without any changes in config or something else...
Cards is RX 580 8Gb.
Are you mining Heavy? If that's the case then the behavior is existent on multiple cards. Yes, heavy. But case is not in cars. Case is in miner. I get similar behavior on xmr-stak on Heavy. xmr stak, gateless gate also have this 'problem'. Maybe this means the 'problem' is in the algo, not in miners? What about compiling kernels every launching? |
|
|
|
because on CPU cryptonight is latency limited.
it means the performance key is all in the cache. since CN Heavy requires twice the cache, you can compute twice less. the compute itself is also slightly slower, but not that much. it makes a Division, which is a bit slow, nothing more. its round does two memory access instead of one, but it makes half of them.
on gpu latency is bad because we do a lot of random access to glogal vram. a gpu can take a whole second to make a cryptonight round. so cache is negligible, and Heavy impact is very low too
you can see that with the no-cache mode of JCE which is slow as hell (often 3 h/s) but not really slower on Heavy than v7.
about gpu, i'm now working on releasing the gpu version of jce. maybe i'll release a last cpu-only then next will be gpu+cpu.
i'll still provide a cpu-only for those who don't have any opencl gpu.
It's good news about gpu-version. I'll wait them. Claymore drops dev of CN miner. GGS is in development. SRB-miner is not stable still. Hope your miner will be good! |
|
|
|
|
zawawa, don't forget to add cryptonight-heavy and lyra2z to nicehash mode. In 1.3.8 i can't connect to this algo...
After pressing "Relaunch" GGS just closes.
|
|
|
|
On AMD 18.5.1 driver on 1803 Windows 10 x64 my RX 580 recognized by SRB miner as R9 200 series [Ellesmere] and every launch build new kernels... It's abnormal. I think, you must update miner to support new drivers, doctor...
By the word, all Claymore's miners detects my cards right on 18.5.1...
And I still can't understand, why every lounch of SRB miner speed is different from each other... To reach max speed it's need to launch SRB several times without any changes in config or something else...
Cards is RX 580 8Gb.
Are you mining Heavy? If that's the case then the behavior is existent on multiple cards. Yes, heavy. But case is not in cards. Case is in miner. |
|
|
|
|
On AMD 18.5.1 driver on 1803 Windows 10 x64 my RX 580 recognized by SRB miner as R9 200 series [Ellesmere] and every launch build new kernels... It's abnormal. I think, you must update miner to support new drivers, doctor...
By the word, all Claymore's miners detects my cards right on 18.5.1...
And I still can't understand, why every lounch of SRB miner speed is different from each other... To reach max speed it's need to launch SRB several times without any changes in config or something else...
Cards is RX 580 8Gb.
|
|
|
|
|
Why on CPU heavy algo is twice lower than V7? On GPU's it's use more vmem but speed is the same or higher...
|
|
|
|
|
Show Posts
